🧠 AI Engineer

🏢 The Company

An early-stage French startup that is developing AI for accounting firms.

🎯 The Role

Build AI features that solve real business problems, from design through to production deployment.

⚠️ What You Need to Know

This isn't a Software Engineer role that dabbles in AI on the side. The team has already ruled out candidates with very strong development backgrounds because they lacked expertise in data and AI—specifically, inference infrastructure, model evaluation, and observability. That's where the selection process really comes into play.

The accounting profession does not tolerate approximations: a plausible but incorrect answer is a mistake. The reliability and traceability of answers take precedence over an impressive demonstration.

🔨 Technical Challenges

• Running RAG on heterogeneous, poorly structured data specific to each firm
• Evaluating an LLM System Thoroughly: Understanding What You're Measuring, Detecting Regression Before the Client Does
• Orchestrate agents in production, along with the monitoring and safeguards that go along with it
• Scale up to portfolios containing several hundred cases that are analyzed continuously

🧑‍💻 The stack

• Backend: Python, FastAPI, pytest
• AI: LLMs, RAG, pgvector / Qdrant, LangChain, LangGraph, LangFuse
• Web and Data: TypeScript, Supabase
• Infrastructure: AWS, Vercel, Docker, CI/CD

🎯 Ideal Candidate Profile

• Experience: At least 3 to 5 years, with a significant portion of that experience working on AI products that have actually been deployed in production, not just proof-of-concepts
• Profile: Hybrid AI Engineer / Software Engineer. Clean, thoroughly tested Python code and a strong background in applied AI
• Expertise: prompting, orchestration, evaluation, vector-based models, model deployment, and monitoring
• Attitude: independent, pragmatic, comfortable with the ambiguity of a team where everything is still to be figured out
• Bonus: prior experience in accounting, finance, or law

🏢 The Work Environment

• 3 days a week on-site in central Paris
• A 10-person tech team: minimal bureaucracy, quick decisions, and a direct impact on the product
• Direct communication with the founders regarding technical decisions

🛑 This job isn't for you if...

• Your AI experience is limited to POCs and notebooks
• You've never had to evaluate or monitor an LLM system in production
• Are you looking for a structured framework, a set roadmap, and ready-made specs?
• Do you want to work fully remotely or less than 3 days on-site?

✅ This job is for you if...

• You've already brought an RAG agent or pipeline into production, and you can explain how it made a difference, backed up by numbers
• You're just as comfortable working with inference infrastructure as you are with application code
• A focus on reliability motivates you more than the rush to build a demo
• Do you want to influence the technical decisions for a product that's still in its early stages?

📌 Prerequisites

• Fluent in French: The team works in French, and the product is aimed at French law firms, whose industry-specific terminology is central to the role
• Valid authorization to work in France, with no action required on the part of the employer
• Based in the Île-de-France region: 3 days a week on-site in central Paris. Applications requiring fully remote work or relocation will not be considered.

🪜 The Hiring Process

• 3 interviews